Trace summarization
Contents
Summarization is currently in beta. We'd love to hear your feedback as we develop this feature.
Trace summarization uses AI to generate human-readable summaries of your LLM traces and events. This helps you quickly understand complex multi-step AI interactions without reading through raw inputs and outputs.
How it works
When viewing a trace or generation event, click the Summary tab to generate an AI-powered summary. The summary includes:
- Title: A brief description of what the trace accomplished
- Flow diagram: An ASCII visualization of the execution flow
- Summary points: Key highlights and actions from the trace
- Interesting notes: Notable observations like errors or unusual patterns
Summarization modes
Choose between two summarization modes based on your needs:
| Mode | Description | Best for |
|---|---|---|
| Minimal | Quick 3-5 bullet points with key highlights | Fast overview of what happened |
| Detailed | Comprehensive 5-10 points with full context | Deep understanding of complex traces |
Requirements
AI data analysis consent
Summarization requires AI data processing to be enabled for your organization. When you first use the feature, you'll be prompted to approve AI data processing. This consent applies organization-wide.
To manage this setting, go to Settings → Organization → General → PostHog AI data analysis.
Rate limits
To ensure fair usage, summarization has the following rate limits:
| Limit | Value |
|---|---|
| Burst | 50 requests/minute |
| Sustained | 200 requests/hour |
| Daily cap | 500 requests/day |
Summaries are cached, so regenerating the same trace won't count against your limits unless you explicitly request a refresh.
Providing feedback
After generating a summary, you can rate it using the thumbs up/down buttons. This feedback helps us improve the summarization quality.